Intelligent Systems and Robotics

Institute of Computer Science — Department VI

In an increasingly competitive world, computers are no longer seen as simple calculating machines, but as a central means of providing intelligent services in business, society and private life. From stock market forecasts to intelligent factory control, from product recommendations to strategic corporate planning, from property monitoring to autonomous lawnmowers: intelligent systems that are able to analyze digital information and sensory data, draw the right conclusions and take action in the digital or physical world are rapidly gaining in importance.

As more and more data becomes available for analysis and decision-making — from web documents and digital media to sensory data from cameras, microphones and ubiquitous devices — intelligent services and support systems of a completely new kind can be created, and autonomous agents and personal robots are on the horizon.

In Department VI “Intelligent Systems and Robotics”, we are working intensively on the development of powerful, robust cognitive and humanoid robots. They are the next step in the fusion of machines, computers, sensors, and software to create intelligent systems capable of interacting with the complexities of the real world. They are the physical embodiment of machine intelligence. Another research focus of our department is the development of techniques that enable robots (operating in human environments) to adapt their behavior to the surroundings by using semantic information about objects and information about users' activities.

Research Highlights

Eine Wissenschaftlerin und ein Wissenschaftler arbeiten hinter einer Glasfassade und mischen Chemikalien mit Großgeräten.

Robotics Meets Cultural Heritage

We are contributing perception and motion planning techniques to the "RePAIR" project, which aims to physically reconstruct destroyed frescoes from Pompeii. Using modern algorithms, we are helping to place fragments with two robotic arms.

Learning Safe and Personalized Robot Navigation

We are researching new navigation algorithms for personalized robot behavior that are easy and safe to learn.

Interactive Perception in Confusing Environments

It is difficult to find objects in confusing scenarios. We are researching algorithms to keep an overview.

Humanoid Soccer Robots with World Cup Title

In the future, robots that can move quickly and robustly on two legs will be important for numerous applications in everyday environments, e.g. for delivering mail or supporting people in need of assistance. Humanoid soccer robots combine precisely these capabilities.

NimbRo: Avatar Robot

Robotic avatar systems provide humans with visual, auditory and tactile information, enabling them to perform tasks in real time over long distances. Avatar technology is seen as the key to enabling people to use their skills in remote locations in various difficult or critical situations, e.g. in care or disaster relief as well as maintenance and repair work.

Working Groups

Department VI "Intelligent Systems and Robotics" consists of two working groups, which are under the direction of Prof. Dr. Sven Behnke and Prof. Dr. Maren Bennewitz.

The "Autonomous Intelligent Systems" group conducts research in the fields of cognitive robotics and computer vision. The main challenges in the design of these robots include the systematic handling of uncertainties, modeling the state of the environment, coordinating teams of cooperating robots in dynamic environments, interaction with humans, development and learning. Examples of cognitive robotic systems that the group is researching include humanoid robots, autonomous micro air vehicles (MAVs), robotic manipulators for waste collection, avatar robots and exploration robots. In the field of computer vision, the working group investigates the categorization of objects, the segmentation of object classes, semantic assignment and action recognition, focusing in particular on deep learning methods, hierarchical representations in neural networks or graphical models.

The research focus of the "Humanoid Robots" group, headed by Prof. Bennewitz, is on robots that act in human environments. The group develops techniques that enable robots to adapt their behavior to the environment and the people around them by using semantic information about objects and information about the user's activities. The main focus is on motion planning and navigation through cluttered and dynamic scenes as well as on the generation of predictive robot behavior by predicting human movements. Another research focus of the group is on active perception, such as the detection of indoor objects or plants in greenhouses with the sensors of a robot to learn 3D models over time.

The third research group, “Robot Perception and Learning”, is headed by Junior Professor Dr. Hermann Blum. It investigates robots that can understand their environment semantically and geometrically in order to carry out manipulations and other safety-critical tasks in the vicinity of humans. In particular, the lab is researching how robots can familiarize themselves with their environment and adapt to changes. This includes recognizing unknown, open objects in the environment, adapting robots to new environments and continuous self-supervised learning. For example, a warehouse robot that learns over time where which item belongs and can adapt independently if the packaging of a product changes.